model the bias

Terms from Statistics for HCI: Making Sense of Quantitative Data

A calculation based on measurements may have bias, that is it may on average either over- or underestimate the real world effect that it is trying to estimate. Sometimes it is possible to work out how large this bias is and then correct the calculation to take that into account. For example, imagine you have a ladder with rungs 20cm apart. You throw a ball up towards the ladder, record which pair of rungs the ball falls between, then multiply the rung it goes over by 0.20 to give the ball height in metres. This will typically underestimate the height of the ball – it is a {biased estimate}}. However, we can model the bias: it is clear that the underestimation is about half the distance between rungs. Having modelled the bias, we can add 10cm to the original calculation and obtain a fair estimate of the ball height.

Used on page 31